Andreas Vlahinos Ken Kelly - AES · Andreas Vlahinos Principal, ... Fundamentals of Designs Of Experiments ... 2007 PTC USER/ WORLD EVENT 35 Achieving Design Requirements

1

Design Space Exploration with

Design Space Exploration with

Behavioral Modeling Behavioral Modeling

Andreas VlahinosAndreas VlahinosPrincipal, Advanced Engineering Solutions, LLCPrincipal, Advanced Engineering Solutions, LLC

PTC USER/ WORLD EVENTNovember 5, 2007 2

Senior Engineer, National Renewable Energy LabSenior Engineer, National Renewable Energy LabKen KellyKen Kelly

Principal, Advanced Engineering Solutions, LLCand

Principal, Advanced Engineering Solutions, LLCand

2

AcknowledgmentsAcknowledgments

� This research effort was supported by the Department of Energy (DOE) Office of the FreedomCAR and Vehicle

� This research effort was supported by the Department of Energy (DOE) Office of the FreedomCAR and VehicleEnergy (DOE), Office of the FreedomCAR and Vehicle Technology. We would like to express our appreciation to:– Robert Kost, team leader of the FreedomCAR and

Vehicle Technology office– Lee Slezak, Technology Manager of FreedomCAR

and Vehicle Technologies Program

Energy (DOE), Office of the FreedomCAR and Vehicle Technology. We would like to express our appreciation to:– Robert Kost, team leader of the FreedomCAR and

Vehicle Technology office– Lee Slezak, Technology Manager of FreedomCAR

and Vehicle Technologies Program


g g– Terry Penney and Keith Wipke of NREL– Pat Davis and Kathi Epping of the Hydrogen, Fuel

Cells & Infrastructure Technologies Program

g g– Terry Penney and Keith Wipke of NREL– Pat Davis and Kathi Epping of the Hydrogen, Fuel

Cells & Infrastructure Technologies Program

BMX what is that?BMX what is that?Pro/ENGINEER Behavioral Modeling Extension (BMX) is a design space exploration and optimizationsoftware that :1. Defines analysis features

• Pro/E measures (distance, curvature, volume, UDA)

• Excel (drag and drop bi-directional associativity)• MDX measures (max acceleration)• MDO measures (max reaction)


• MDO measures (max reaction)• External programs (FEA, CFD, Cost, etc)

3

BMX what is that?BMX what is that?• Pro/ENGINEER Behavioral Modeling Extension (BMX) is a

design space exploration and optimization software:

• Using these analysis features perform:• Sensitivity Analysis• Feasibility / Optimization• Multi-objective Design

Studies


• Generates optimization features that extend the Pro/E associativity to product attributes and drives the design from engineering requirements (not dimensions)

Tools for Robust DesignTools for Robust Design� Design Of Experiments

– Exploits nonlinearities and interactions between noise & control parameters to reduce product performance variability

� Design Of Experiments– Exploits nonlinearities and interactions

between noise & control parameters to reduce product performance variabilityp p y

– full factorial, fractional factorial, Monte-Carlo, LHC

� Response Surface Methods– Central Composite Design– Box-Behnken Design

6 sigma design

p p y– full factorial, fractional factorial, Monte-Carlo,

LHC

� Response Surface Methods– Central Composite Design– Box-Behnken Design

6 sigma design


� 6-sigma design– Identifying & qualifying causes of variation– Centering performance on specification target– Achieving Six Sigma level robustness on the

key product performance characteristics with respect to the quantified variation

� 6-sigma design– Identifying & qualifying causes of variation– Centering performance on specification target– Achieving Six Sigma level robustness on the

key product performance characteristics with respect to the quantified variation

4

Design Exploration 1 VariableDesign Exploration 1 Variable

1.5

-0.5

0

0.5

1


0 10 20 30 40 50 60 70 80 90 100-1.5

-1

Design Exploration 2 VariablesDesign Exploration 2 Variables

0.9

1

0.3

0.4

0.5

0.6

0.7

0.8


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

5

Design Exploration 3 VariablesDesign Exploration 3 Variables


Example - Live DemonstrationFundamentals of Designs Of ExperimentsExample - Live DemonstrationFundamentals of Designs Of Experiments

Problem Statement:

Find the lightest section that meets the Design

Problem Statement:

Find the lightest section that meets the Designmeets the Design requirements (Ixx > Ireq) for the aluminum extruded section shown in 5 min.

meets the Design requirements (Ixx > Ireq) for the aluminum extruded section shown in 5 min.


6

Problem Statement in ProE termsProblem Statement in ProE terms

� Find the dimensions (Pro/E parameters W, H, t ) that minimize the cross sectional area (Min W i h ) d ll h h

� Find the dimensions (Pro/E parameters W, H, t ) that minimize the cross sectional area (Min W i h ) d ll h hWeight - cost) and meet all the strength, manufacturing and stability requirements80 < W < 120160 < H < 2404 < t < 6

Weight - cost) and meet all the strength, manufacturing and stability requirements80 < W < 120160 < H < 2404 < t < 6


� Action: Perform a multi-objective design study to identify the most economical section that satisfies the strength requirements.

� Action: Perform a multi-objective design study to identify the most economical section that satisfies the strength requirements.

Behavioral Modeling of the Rocker SectionBehavioral Modeling of the Rocker Section

•All sections have the same moment of Inertia

•Find the one that minimizes the cross sectional area (Min Weight) and meets all the manufacturing and stability requirements


manufacturing and stability requirements

•Not a dimension driven CAD model

•Requirement driven design (Ireq)

7

Typical Design Exploration (without BMX)Full Factorial, 2 level - 3 factor designTypical Design Exploration (without BMX)Full Factorial, 2 level - 3 factor design

Normalized Design Space

1

0

1

t


-1

0

1

-1

0

1

-1

W H

Response Table for Three-factor ExperimentResponse Table for Three-factor Experiment

min 80 max 120 min 160 max 240 min 4 max 6

t (thickness)Experiment Number

Response Value

W (Width) H (Height)

1 R1 R1 R1 R1

2 R2 R2 R2 R2

3 R3 R3 R3 R3

4 R4 R4 R4 R4

5 R5 R5 R5 R5

6 R6 R6 R6 R6


6 R6 R6 R6 R6

7 R7 R7 R7 R7

8 R8 R8 R8 R8

AVERAGE R W1 W2 H1 H2 t1 t2

EFFECT W2-W1 H2-H1 t2-t1

8

Graphical Representation of Main EffectsGraphical Representation of Main Effects

5500

6000 Graphical Representation of main effects

4000

4500

5000

Res

pons

e V

alue


1 2 32500

3000

3500

Parameter

R

Design Exploration with BMXScatter Plot for two of Design VariablesDesign Exploration with BMXScatter Plot for two of Design Variables


9

Scatter Plot of Ixx versus AreaScatter Plot of Ixx versus Area

gth


cost

Stre

n

Live DemonstrationLive Demonstration


10

Chassis Design ProcessChassis Design Process

Manufacturing Supplier Guidelines

t>2, R>1

Shared Knowledge

(Local Buckling, tf/tw)

Structural Performance

Targets (m,Kt,Kb,ω)

Economic AnalysisCost (material,

fabrication, assembly)

StylingPackaging

BehavioralModeling

Beam AnalysisDesign of

Experiments

Optimum Beam SectionsOptimum Connection Shape

TopologyOptimization


Optimum Connection Shape Optimization

CAD DesignBody in White FEA / Optimization

Body in White Final ChassisCAD Design

MeetTarget

BMX Example on Power Electronics CoolingBMX Example on Power Electronics Cooling

Project Goal : Develop a heat exchanger design


Develop a heat exchanger design to efficiently remove heat from the power module and reject it into the vehicles coolant loop with uniform cooling, minimum cost, volume and pressure drop

11

Problem StatementProblem StatementF find the optimal pin-fin geometry that:

Minimizes dT

Where:

F find the optimal pin-fin geometry that:

Minimizes dT

Where:Pin h

Pin dia

1 mm < Pin_dia < 10 mm

1 mm < Pin_h < 5 mm

1 < Nx < 15 (integer)

2 < Ny < 50 (integer)

Subject to:

1 mm < Pin_dia < 10 mm

1 mm < Pin_h < 5 mm

1 < Nx < 15 (integer)

2 < Ny < 50 (integer)

Subject to:


Subject to:maxT < 125 °C

dP < 20000 Pa

(Lx-Nx*Pin_dia)/Nx > 0.5 mm (no interference in x)

(Ly-2*Ny*Pin_dia)/(2*Ny) > 0.5 mm (no interference in y)

Subject to:maxT < 125 °C

dP < 20000 Pa

(Lx-Nx*Pin_dia)/Nx > 0.5 mm (no interference in x)

(Ly-2*Ny*Pin_dia)/(2*Ny) > 0.5 mm (no interference in y)

Ny

Nx

Design Aids: Experiments, CFD, FEA, BMX, CBDesign Aids: Experiments, CFD, FEA, BMX, CB


12

CFD Analysis of Staggered vs. In-Line Flow


Thermal AnalysisClassical Determination of Film Coefficients for BMX ImplementationThermal AnalysisClassical Determination of Film Coefficients for BMX Implementation

Reynold’s number for flow around pin fins

Nusselt numberl i

36.0Pr4.0Re90Nu =

ν

DU

DmaxRe =

Heat transfer

- laminar

- transitional

- turbulent

PrRe9.0DD

Nu =

36.0Pr6.0Re2.0

35.0DSl

StD

Nu ⎟⎠⎞

⎜⎝⎛=

36.0Pr84.0Re022.0DD

Nu =

kNu ×


coefficient

Pressure Drop N

UfP

2

2maxρ

=Δ

DkNu

h×

=

13

Feasible Design Space of Ny versus Nx

Pin dia

Pin h


Ny

Nx

Feasible Design Space of Pin diameter versus NxFeasible Design Space of Pin diameter versus Nx


14

Design Option A


Design Option B


15

Design Option C


Design Option D


16

Design Option E


Temperature Distribution with BMX’s UDATemperature Distribution with BMX’s UDA


17

200

250

dP Limit

Design Space Exploration (pin h, pin dia, spacing)Maximum Temperature vs. Pressure Drop

100

150

Tmax

(deg

C)

Pin h = 0.001 m

Pin h = 0.003 m

Pin h = 0 005 mT t

Tmax Limit


0

50

0 10000 20000 30000 40000 50000 60000 70000 80000 90000 100000

Pressure Drop (Pa)

Pin h = 0.005 mTargetDesignSpace

50

60

dP Limit

Design Space Exploration (pin h, pin dia, spacing)Temperature Differential vs. Pressure Drop

20

30

40

dT (d

eg C

)

Pi h 0 001

dT Limit


0

10

0 10000 20000 30000 40000 50000 60000 70000 80000 90000 100000

Pressure Drop (Pa)

Pin h = 0.001 m

Pin h = 0.003 m

Pin h = 0.005 m

Target Design Space

18

Identifying the Best Solution


Achieving Design Requirementswithin the CAD Environment

Attribute driven Parametric modeling (dP, Tmax, dT)

Automated optimization atAutomated optimization at the design stage

Very fast solutions and flexible geometry

Requires closed form


solutions or link to other analysis tool (CFD, FEA, etc.)

19

Behavioral Modelinga paradigm shift in the design process

“The problem is never how to get new innovative thoughts into your mind, but how to get the old ones out”


Documents

Andreas Vlahinos Ken Kelly - AES · Andreas Vlahinos Principal, ... Fundamentals of Designs Of Experiments ... 2007 PTC USER/ WORLD EVENT 35 Achieving Design Requirements